May Bubble or Overblown: Ambiguous Signals in the Subprime Auto Market. By Robbin Conner

Transcription

1 May 2016 Bubble or Overblown: Ambiguous Signals in the Subprime Auto Market By Robbin Conner

2 Contents SUMMARY...3 MARKET SIZE...4 CREDIT QUALITY OF SUB-PRIME ORIGINATION...4 STABILITY OF RATINGS...7 OTHER FACTORS...8 TERM OF BORROWINGS...9 SUBPRIME LOAN AMOUNTS DELINQUENCIES ADDITIONAL FACTORS CONCLUSION ABOUT THE AUTHOR

3 Summary In the aftermath of the great recession the Federal Reserve has overseen a lengthy period of low rates, designed to aid recovery. However, such an accommodative policy risks creating a bubble of artificially inflated asset values that could sow the seeds for the next wave of distress. One sector drawing attention in this regard is subprime auto ABS, which has seen a resurgence in lending to consumers with poor credit scores. With investors in need of yield and the persistence of a generally favorable credit environment since the end of the recession, the question is whether the pump is being primed for a strong correction in this sector or whether the lessons of the last recession have been absorbed sufficiently to preclude an unpleasant recapitulation of events in the sub-prime mortgage market. At the moment the evidence, while not alarming, is somewhat ambiguous and warrants consideration. Published articles 1 have come down on both sides of the issue. On the negative side the articles typically focus on the increasing amount of subprime debt available, the extension of the loan terms, and rising delinquencies in a period of low stress. On the positive side are the continued strong ratings, the amount of credit enhancement in many transactions, and a recognition that while subprime lending has increased relative to post-crisis years, the subprime percentage is merely normalizing after a period of pent-up demand. On examination, many of the negative issues seem less troublesome than portrayed. On the other hand, other factors such as the rising rate of delinquency and loss in subprime pools, particularly in this benign economic climate, indicate legitimate reasons for caution. As we will describe herein, to the extent the press has focused on the growth of subprime lending, it has typically done so by reporting growth in subprime loans on an absolute rather than a relative dollar basis; singling out particular deep subprime originators such as Skopos or Drive; or portraying growth relative to the narrow post-crisis period, when subprime lending had been unusually curtailed. 1 See for example, Zero Hedge Subprime Auto Goes Full Retard: Lender Sells $154 Million ABS Deal backed By Loans to Borrowers with No Credit Tyler Durden, November 3, 2015; Bloomberg Longer Leash for Subprime Car Buyers in U.S. Stokes Debt Concern Matt Scully, June 23, 2015, 8:21 PM EDT, updated June 24, 2015, 12:50 PM EDT; Center for Responsible Lending Reckless Driving : Implications of Recent Subprime Auto Finance Growth January 2015; NY Times, DealB%K Investment Riches Built on Subprime Auto Loans to Poor Michael Corkery and Jessica Liver-Greenberg, January 26, 2015; David Stockman s Contra Corner Here They Go Again Subprime Delinquencies Rising in Autoland David Stockman, January 11, 2015; mybudget 360 How Subprime loans keep the bubble going: Subprime auto loans continue to grow as credit worthy customers drop out of the market. September But see, for example, The Wall Street Journal Why Fears of a Subprime Auto Bubble Are Overblown for Now Nick Timiraos, September 24,

4 Billions Market Size Without question, the size of the auto loan market together with subprime lending has expanded. Outstanding debt for the auto loan market as a whole is now over $1 trillion, according the Fed s most recent report. Credit Quality of Sub-Prime Origination Some articles have raised concerns about the loosening of credit standards in subprime auto lending. Certainly, in particular platforms that have become active after the crisis, there has been a focus to lend to lower-tier borrowers and extend $1.2 $1.0 $0.8 $0.6 $0.4 $0.2 Auto Loans $ Source: The Federal Reserve Bank of New York, Quarterly Report on Household Debt and Credit February 2016 the term of financing. For example, Skopos, a private equity backed entity that opened in 2012 specializing in very deep subprime often geographically concentrated collateral was able to grow its portfolio enough to issue its first rated deal at the end of 2015, albeit from a still relatively small pool of $157MM of receivables; and, after a long hiatus, Santander has brought back its Drive platform specializing in deep subprime paper. The collateral characteristics of Skopos-originated loans are in line with the reports of expansion of PE-backed companies plunging into deeper subprime territory. Borrowers with either no FICO score or FICO scores below 500 account for more than 30% of the pool, and only about 10% have FICOs above 600, with the weighted average FICO being about 540. The weighted average original term is 70 months, with two-thirds of the pool financing used vehicles. Experian reported that for all deep subprime borrowers the average term was 58 months for used vehicles and 72 months for new vehicles. Although Skopos did not have large portfolio growth prior to 2014, its early vintage losses from smaller portfolios originated in 2013 have already hit 25%, a high number in benign times, and equivalent to or higher than many subprime lenders experienced during the crisis. Finally, the concentration in Texas is unusually large at almost 50%, and it remains to be seen whether this group will be affected by regional factors connected to the oil patch. These levels are comparable, in some respects, to those of Santander s relaunched Drive platform, which issued securities backed by a $1.3B pool a month and a half earlier. Like Skopos pool, roughly 30% of the pool had either no FICO or a FICO 500. Unlike Skopos, it had 37% (vs 27%) in the higher band and a lower concentration (17.5% vs 31%) in the riskier band. Like Skopos, about 90% of the loans had an original term of about 70 months and the percentage of new vehicles was even lower at 29% (vs 33%). Its largest concentration is also in Texas but represents only 11% of the pool. 4

5 Billions While these are two examples of platforms focused on deep subprime that have entered the market post crisis, they are not representative of the market as a whole. In fact, the percentage of subprime auto loans being made is less than pre-crisis. This longer-term pattern is evident in the figure below. Equifax Scores < 620: The percentage of loan originations with Equifax Scores < 620 dropped to a low of about 17% in 2010 before rising, based on yearly averages, to approximately 22% between 2012 and 2014, and then increasing slightly (but still to under 23%) in This level is still noticeably below not only the high water marks set in 2006 and 2007 but also the roughly 26% average levels of 2004 and 35% 30% 25% 20% 15% 10% 5% Lower Tier Risk Scores as a Percentage of Originations While the data is not granular below 620, so any deterioration in the distribution of scores below this threshold would not be visible, the data alone does not presage any dramatic increase in risk; rather, it appears to show a return to something closer to long term historical norms. < % Source: The Federal Reserve Bank of New York, Quarterly Report on Household Debt and Credit February 2016 Equifax Scores : The pattern is somewhat similar for loans with Equifax scores from , which declined postcrisis and reached an average low in 2009 of about 11%, rose to average levels of 13.7% in 2012 and 2013, and then declined slightly to a 13.2% average in 2014 and 2015 a level very slightly higher than 2004 but lower than Again, this increase from 2009 levels is not alarming when viewed within its historical context. 2 $1,000 $900 $800 $700 $600 $500 $400 $300 $200 $100 $0 Subprime Outstandings and Percentages Total Outstandings Deep Subprime Outstandings Deep Subprime Percentages 16.9% 16.6% 16.8% 3.8% 3.8% Subprime Outstanding Subprime Percentages % Source: The Federal Reserve Bank of New York, Quarterly Report on Household Debt and Credit February % 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% 2 Our analysis assumes no change in meaning or scale of the risk score during this time. 5

6 Experian s data on outstandings, while slightly different, paints a similar picture over the past three years. It is worth reiterating that, as evident in Experian s data as well, while outstandings have grown 23% since the 4 th quarter of 2013, the subprime exposure has been stable, comprising 16.9%, 16.6% and 16.8% of outstandings in the fourth quarters of 2013, 2014 and 2015, respectively. The exposure to deep subprime borrowers has been similarly stable, with percentages of 3.8%, 3.8% and 4.0% in the same years. The question is therefore one of scale. Are there sufficient numbers of subprime borrowers that will behave similarly to earlier borrowers (with comparable risk scores) to accommodate this rate of growth? In short, is there an adverse selection risk that is not picked up by credit scores as the market continues to expand? Only time will tell. However, none of the growth metrics is showing a worrying concentration of risk in the subprime space only a growth rate proportional to outstandings in all classes of auto loans. Looking at the entire distribution of scores on an annual basis shows potentially positive signs at the higher end of the distribution. While the percentage of loans with 780+ Equifax scores has declined since the years immediately following the crisis, it has been stable since 2013 and noticeably above the levels seen in 2004 and The following chart highlights these differences, comparing the loans originated in the 4 th quarter of 2015 to those originated in the 4 th quarter of prior years. It subtracts the percentage value of 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Yearly, Q4, Originations by Equifax Score < Source: The Federal Reserve Bank of New York, Quarterly Report on Household Debt and Credit February 2016 loans originated in the prior periods from the percentage originated in For example, if the percentage of loans with risk scores of less than 620 originated in the 4 th quarter of 2015 was 21.8% and the percentage of loans with Equifax Scores of less than 620 originated in the 4 quarter of 2004 was 27.4%, the graph would show a negative 5.6% value that is to say that the distribution of 4 th quarter scores in 2015 has 5.6% fewer loans with risk scores of less than 620 than that of the 4 th quarter As can be seen, the distribution of borrowers credit scores has been largely stable in 2013, 2014 and If anything, starting in 2013 there was a slight increase in scores between 660 and 719, with decreases in scores below 620. In addition, in 2015 more loans were underwritten at the very high end of the scale, 780+, with significant decreases in loan with scores below

7 Relative Differences between Origination Percentages in Q vs. Q4 of Past Years 10% 8% 6% 4% < % 0% -2% -4% -6% -8% Source: The Federal Reserve Bank of New York, Quarterly Report on Household Debt and Credit February 2016 If the floodgates are being opened to poorer credits in the search for yield, it is not evident in the credit score data. Stability of Ratings Another area of strength so far is the stability of ratings. While delinquencies and losses are increasing, S&P has recently published a complete review of the subprime sector 3 and did not find any trends that are likely to impact ratings this year. Explaining their rationale, they write, The aforementioned increase in losses continue to be mitigated by credit enhancement levels and robust structural features. As a result, we expect ratings to remain stable this year... In short, in their rated transactions, they believe the relatively high levels of hard enhancement and excess spread will counteract any increase in collateral losses even with respect to the issuers specializing in the subprime and deeper subprime spaces. However, the revisions in current expected pool loss, relative to the original loss expectation at the time of rating, vary significantly by trust. Some, such as Americredit, have actually lowered cumulative pool loss expectations or, like Santander s SDART program, have stable estimates with generally no increased expectation of loss. Others, such as Prestige Auto Receivables Trust, CPS Auto Receivables Trust, or United Auto Credit Securitization Trust, have a noticeable increase in percentage terms of cumulative net losses. As can be seen in the table below S&P has increased the expected range of loss on the United Auto Credit by 22%; the loss expectation for certain CPS Trusts by 24%; and the loss expectation for the Trust of Prestige by 58%. 3 Standard and Poor s Ratings Services, U.S. Subprime Auto Loan ABS Tracker: December 2015 February 26,

8 Trust Original Loss Range Revised Loss Range Change United Auto Credit % 20%-21% 22% CPS 13%-13.5% 16.25%-15.75% 24% Prestige %-7.4% 11.25%-11.75% 58% While there is a significant difference in change in expected losses among particular programs that may be of interest to investors considering those programs, none of the changes is at this time, according to S&P, enough to warrant an S&P rating action or to elevate downgrade concerns generally. Other Factors However, despite the reasonable distribution of subprime originations and the confidence of S&P in the current structures and protections, there are still some significant negative issues often raised that deserve attention. These issues include the increase in the term of the borrowings, a growth in the amounts being financed, and the deterioration in performance metrics that seem difficult to explain in this currently benign credit environment. There has been some increase in the term of loans, although the increase is much more significant for prime than for subprime borrowers but that trend, has either slowed or, in the case used-vehicle loans, reversed slightly (relative to last year within the subprime FICO bands). Consequently, despite the potential negative effects associated with lengthening the term of loans extended, the changes with respect to subprime borrowers do not appear dramatic and the current rate of change is somewhat favorable. Similarly, the amounts financed have increased and could affect both frequency and severity of loss but that increase needs to be viewed in terms of real dollars as well as the specific increase in prices for cars. When adjusted for inflation, average loan amounts to borrowers in the deepest subprime band have actually decreased and the loan amounts to the remaining subprime borrowers do not appear as large as often reported. Finally the rise in delinquencies is real, however. If credit scores, lengthier loan terms, and increased loan amounts do not explain this rise of delinquencies that is occurring even in a moderate credit environment, one has to ask what does. It also raises the question of whether such weakness now portends more extreme losses when the cycle turns than were experienced in the last period of stress. 8

9 Term of Borrowings Experian data shows a modest increase in the term of subprime new-car loans since 2008, with the average term rising from 68 months in 2008 to 71 months through August The pattern for used cars shows a drop in average term from 57 months in 2008 to 52 months in 2009, followed by increases to 60 months by 2014 a different pattern, with sharper interim decreases and increases in particular years, but also resulting in a net increase of 3 months from 2008 levels. In addition, this data does not show the composition of the average on a more granular basis. Experian recently published its 4Q2015 review that shows why this granularity could be useful. Comparing 4 th quarters of 2015 to 2014 it noted an overall increase (not simply for subprime loans) of loans for both new and used vehicles in the month range and more sharply in the month ranges. For new cars it noted that 42% of loans had terms of months while 29% of loans now had terms of months, year on year increases of 5% and 12% respectively. For used vehicles, 41.1% had terms of months while 16.4% had terms of months, reflecting year on year increases of 2.6% and 10.8%, respectively. New Subprime Loan Terms Used Source: Experian, State of the Automotive Finance Market Fourth Quarter 2015 Melinda Zabritski Q vs Q Subprime Loan Term New Subprime New Deep Subprime Used Subprime Q4 14 Q4 15 Used Deep Subprime Source: Experian, State of the Automotive Finance Market Fourth Quarter 2015 Melinda Zabritski Interestingly, these average increases were not reflected in the terms to subprime borrowers in the quarter-to-quarter comparison. For subprime borrowers financing new cars, terms increased marginally from 71.6 months to months and for deep subprime borrowers, term increased even less from months to months. For subprime and deep subprime borrowers financing used cars, average term decreased from to months and from to month respectively. Consequently, while there has been a general increase in the term of loans made to subprime borrowers since 2008, the most recent quarterly comparisons suggest that such increases are slowing or, in the case of used vehicles, potentially reversing slightly. Although the trend to increase the term of financing to subprime borrowers certainly merits close scrutiny, it does not appear that originators in the last year have pushed too aggressively on the term front, to keep sales moving in 9

10 the deep subprime space. What effect the earlier, moderate, increases in term have had on pool losses is unclear, but does not appear unduly large. Interestingly, the industry itself has taken divergent views on the extension of terms. Tom Dundan, former Santander Consumer USA CEO, was quoted in a Bloomberg article in July 2015 as saying: The data would tell you that as long as the payment is affordable and quality of the vehicles is improving, the math says you d rather give more term so the customer can get a better car. He also said seven years may soon be the standard for prime loans. In another Bloomberg article (June 23, 2015) Andrew Kang, then of Exeter and now back at Santander, said: At this time we have no intention of going longer than 72 months. The risk is that you extend a loan that a borrower cannot afford over its term schedule. Inching out to 75 and 84 months, I don t think that has been tested yet. Echoing those thoughts, Anup Agarwal of Western Asset Management Co said: Everyone has used the argument that borrowers pay car loans because they have to get to work But borrowers only pay loans if the car is working. We have not seen this cycle come through yet. This potential nuance that may require some thought, although as indicated above the most recent quarter seems to indicate a slowing in term increases offered to subprime borrowers, and overall the increase in term would seem modest over the last several years. Subprime Loan Amounts The size of average borrowings has also been increasing as has the percentage of vehicles with financing. However, again the headline numbers may not tell the whole story. It is true that the average amount financed increased from $24,183 in 2008 to $28,659, a growth of about 19%, by August of 2015 (and for subprime borrowers purchasing new cars from $14,738 to $15,730, or about 7%), but over the same period inflation was about 13%. In real dollars, therefore, used car balances decreased and new car balances experienced modest growth. Delinquencies Subprime Loan Amounts ($000) Delinquencies, however, cannot be so easily explained. According to S&P, whether measured as annualized monthly numbers for a particular month (December) or as annual averages of such annualized monthly rates, delinquencies and losses have been rising noticeably. Starting from 2012, loss rates and delinquencies have New Used Aug Source: Experian, State of the Automotive Finance Market Fourth Quarter 2015 Melinda Zabritski 10

11 increased by 47% and 50% respectively, when measured monthly, or by 44% and 50% when computed from annual numbers. 12% 10% 8% Monthly Annualized Loss and Delinquency Rates Net Loss Rate 60+ Delinquency 12% 10% 8% Annual Averages of Monthly Loss and Delinquency Rates Net Loss Rate 60+ Delinquency 6% 6% 4% 4% 2% 2% 0% Dec Dec Dec Dec Dec Dec Dec % Source: Standard and Poor's Ratings Services U.S. Subprime Auto Loan ABS Tracker December 2015 February 26, 2016 Source: Standard and Poor's Ratings Services U.S. Subprime Auto Loan ABS Tracker December 2015 February 26, 2016 Picking 2011 as the starting point for the annual averages changes the numbers slightly since the net loss rate measured on this basis declined between 2011 and 2012 before rising monotonically but not the major thrust. Computed on this basis, the increase in net loss rate to 2015 is about 29% while the increase in 60+ delinquencies is marginally higher at about 53%. These increases in delinquency are surprising in this benign environment and given the relatively innocuous changes in risk scores, term and amounts borrowed. It augurs caution simply because these measures and the environment do not explain such an increase. Certainly it could suggest higher loss outcomes than previously experienced if the macroeconomic environment were to suffer as it did in the last crisis. Additional Factors Several areas of risk are not mentioned as often, but could explain some of the deterioration in performance. LTV and DTI measures might be deteriorating. It is also possible that given the large absolute amount of lending, certain adverse selection is occurring even with similar risk scores, as deeper pockets of the subprime world are being explored. It would also be useful, although rarely discussed or presented, to know the discount rate at which the servicers are acquiring loan paper from the dealers. This discount rate would be a good indicator of how the companies closest to the risk were evaluating current production compared to earlier times. Conclusion There has been an increasing attention in the press to the potential for a bubble in the subprime auto lending sector, especially given the general hunt for yield in this low interest rate environment and the pressures associated with new market entrants funded by private equity. Casual evidence has been adduced to this claim by showing the growth in subprime lending at low risk scores, the increasing terms of loans and the amounts being financed. However, the data to date on these risk measures, when examined, do not provide cause for concern. The 11

12 percentage of loans being originated to subprime borrowers has increased since the period of tightening immediately after the crisis but remains well within pre-crisis bounds. Loan terms for autos in general have increased, but in the subprime space the increase has been modest, with some recent trends in the deep subprime risk grades actually showing a reduction in term since last year. Adjusting for inflation, the average dollar amount borrowed by subprime borrowers is also unremarkable. However, delinquencies and losses have been steadily increasing since 2012 and, given the stability of the risk measures discussed above, this increase cannot be adequately explained. It is possible other variables would explain the risk but none has been provided to date. While some of the criticism appears overdone, and the scale of subprime lending remains small relative to pre-crisis mortgage originations, some questions remain concerning the state of the subprime market and its potential for elevated losses in another downturn given the rapidly rising loss and delinquency levels in today s unstressed environment. 12

13 About the Author Robbin Conner PF2 Securities Evaluations, Inc. Robbin has more than 20 years of experience with ABS asset classes, including flow asset classes such as credit cards, auto loans and leases, CDOs, small ticket equipment leasing and student loans; a wide variety of esoteric securitizations such as solar leases, time share receivables, servicer advances, peer to peer lending, single family rental, and small box single property CMBS; and most transportation classes such as aircraft, rail and container leasing transactions. His background encompasses leadership and analytic roles on the buy side, sell side and with the rating agencies; and he has led teams analyzing ABS, CDO and mortgage risk domestically and abroad. Most recently Robbin was an MD at Credit Suisse where he ran the asset selection and risk assessment for the bank s $19B ABCP conduit. Previously he was the COO and head of office for a Bermuda fund specializing in synthetic structured risk and prior to that an MD at Moody s Investors Service in structured products in NY and London. He has extensive experience in the risk measurement of ABS tranches, model development, portfolio management, compliance, and foreign bank capital regulations relating to structured funds. Robbin began his ABS career as an attorney at Skadden, Arps, Slate, Meagher & Flom LLP where he practiced for more than 5 years. This, together with his experience on the business side negotiating ABS transactions, has provided a strong technical knowledge of the structures and typical transaction documents employed in most ABS asset classes. Robbin received his J.D. from Northwestern University and a B.A. in economics from Carleton College. 13

14 PF2 SECURITIES EVALUATIONS, INC. 14